1

Auction Experiments and Simulations of Milk Quota Exchanges

Bernhard Brümmer, Jens-Peter Loy and Till Requate1

Paper prepared for presentation at the EAAE 2011 Congress

Change and Uncertainty Challenges for Agriculture, Food and Natural Resources

August 30 to September 2, 2011 ETH Zurich, Zurich, Switzerland

Keywords: Double Auction, Experiment, Milk Quota, Germany

JEL-codes: L11, D44, Q13

Copyright 2010 by Bernhard Brümmer, Jens-Peter Loy, and Till Requate. All rights reserved.

Readers may make verbatim copies of this document for non-commercial purposes by any

means, provided this copyright notice appears on all such copies.

1 Jens-Peter Loy is Professor for Market Analysis at the Department of Agricultural Economics at the Christian-

Albrechts-University in Kiel, Germany; Till Requate is Professor for Innovation, Competition and New

Institutions at the Department of Economics at the Christian-Albrechts-University in Kiel, Germany; Bernhard

Brümmer is Professor for Market Analysis at the Department of Agricultural Economics at the Georg-August-

University in Göttingen, Germany.

2

Abstract

Since 2000 Germany has introduced a fairly unique market mechanism to trade milk quotas

between dairy farms. The two major features are: (1) a quasi auctioning system that produces

excess demands which are covered by state reserves free of charge and (2) a price band that

is used to exclude highest bids. For both features an experimental design is developed to

study the impact in reference to a regular seller’s sealed bid double auction. Results show

that both treatments lead to significant misallocations. These are due to the direct impact of

regulations and due to an imperfect adjustment of bidding functions. The major goal of the

market design to reduce quota prices is reached, however, at significant trade losses.

Introduction

In 1984 the European Union introduced a quota system to limit milk production in the

Member States. Economies of scale and technical progress have set incentives for competitive

farms to grow constantly; thus, there is an inherent need to trade milk quotas between dairy

farms. Quota trades between farms were regulated from the start and differed between EU

Member States till today; however, at the end of the 1990’s many regulations had been lifted

in Germany where in April 2000 a fundamentally new milk quota exchange system was

introduced. A quasi double auction was installed at a regional level. All other forms of

transactions such as renting, leasing or other contracts to transfer quota between farmers were

banned. The new quota exchange system is similar to a single price multiple unit seller’s

sealed bid double auction (SBDA). The SBDA serves here as a reference scenario (Treatment

1). Specific regulations were added to limit prices and to reduce problems of excess supply or

demand. Three major features describe the characteristics of the market mechanism. First, a

single preliminary price is set at the minimum excess demand which is calculated from

ordered bids and asks submitted at each of three auctioning dates per year. Second, if there are

successful bids that exceed the preliminary price by more than 40 percent (and the price is

above 30 Euro cent per liter), then those exceeding bids are cancelled and price calculations

start over again for one more time (Treatment 2). Third, local authorities control some state

reserves of milk quotas. These quotas can used to fill the excess demands that systematically

occur in the auctions. Quotas taken from state reserves are given to bidders free of charge.

Every successful bidder receives the same share with respect to his/her bid volume. Thus, if

the excess demand makes for example 10 percent of the trade volume, then each bidder gets

10 percent of his/her bid free of charge (Treatment 3). We analyze the two scenarios by

comparing their results to the outcome of a SBDA in the first Treatment.

In multiple unit bid auctions with sufficiently many bidders on both sides truth telling is a

weak dominant strategy (Nautz, 1995; Nautz und Wolfram, 1997; McAfee, 1992). The SBDA

auction design severs as a benchmark to analyze the two scenarios for German milk quota

exchanges described above. As the optimal competitive bidding functions under Treatment 2

and 3 cannot be derived easily, we use simulations instead to estimate them. In the second

step auction experiments are set up in a computer lab to study the real world performance of

the market mechanism. Each experiment for the two scenarios is run with 16 student of which

randomly assigned 6 serve as sellers and 10 as buyers of milk quota. Each participant receives

a willingness to buy or sell in each of the 28 rounds played. In each round bids are placed,

prices are calculated, and gains and losses due to transactions performed are assigned and are

notified for each participant. Students receive a fixed and a variable payment which on

average is about 10 Euro per participant. One experiment with 28 rounds and three test runs

takes about 60 minutes. Each design (Treatment 2 and 3) is repeated 5 times with different

students. Willingness to pay and to sell are taken from a uniform distribution between 0 and 1

which fits the scale of real quota prices in Germany. These parameters are also used in the

theoretical simulations to generate optimal bidding functions.

3

The paper is organized as follows: We first describe and analyze the functioning of milk quota

exchanges in Germany. Following, the experimental design for ht three treatments used is

developed and the experimental set up is explained. In section 3 optimal bidding behavior and

expected results are derived and discussed. In section 4 we analyze the experimental results

with respect to the theoretical expectations and measure the variations between the treatments

and between times in the experiment. Finally, we briefly summarize the paper and draw

conclusions.

German Milk Quota Exchanges

While most regulations regarding the milk quota system are set at the EU level, the rules for

the exchanging quotas between farmers are decided at the national level. Consequently the

applied rules differ significantly between member states. After the introduction of the quota

system in 1984 the exchange of quotas between farmers was initially very restrictive. Quotas

could only be transferred when also a specified acreage of grassland was transferred.

However, regulations were liberalized over time and at the end of the millennium farmers

could either buy, rent, or lease quotas without any additional obligations within the regional

territories of dairy companies, in particular only farmers that deliver to the same dairy plant

could exchange quotas. Trade was generally organized by professional quota brokers. In 2000

the German Minister of Agriculture started a new initiative to trade milk quotas through state

organized exchanges that are held three times a year. Instead of limiting exchanges to the

regional territories of the dairy companies, Germany was divided into 21 regions, for which

separate exchanges were set up. The regions are mainly congruent with the state territories.

Only in Bavaria and in Baden-Württemberg certain sub-territories were used the called

Regierungsbezirke. From 2007 on regions are aggregated into East and West Germany, for

which separate exchanges are run. Since 2000 with only a very few exceptions all trades have

to go through the state exchanges. Also renting and leasing of quotas were banned.

The exchanges are organized as quasi double auctions. Bids and asks are submitted by

farmers to the exchanges which then decide a single price and the quantity that is traded. All

bidders have to prove that they can afford to buy quotas (bids) or that they are in possession

of the quota they intend to sell (asks). All bidders can submit only one bid. Trading rules have

been changed a few times between 2000 and 2002. Since then the following procedure is

applied. First, incoming asks and bids are ranked by priority and cumulated. Second, starting

with the minimum bid/ask, the market excess quantity (cumulative bids minus cumulative

asks) is determined for every Eurocent increment up to the maximum bid/ask. The

preliminary price is set where the minimum excess demand is reached for the first time or at

the lowest price. Third, if the preliminary price is lower than 30 Eurocent, then it is equal to

the final price. If the preliminary price is above 30 Eurocent and if there are bids higher than

1.4 times the preliminary price, then those bids are deleted and the price determination starts

again. The second round will always be the final round. If there are no such bids, then again

preliminary and final price correspond. Fourth, after the preliminary price is found and the

excess demand is determined, it is decided whether all bids are cut by the same share with

respect to bid volumes to reach equilibrium (supply equals demand) or whether the state

reserve is used to fill the gap. If the state reserve has a sufficient amount of milk quota, it is

generally used. In this case every bidder receives the same share of the excess demand

according to his/her bid volume free of charge.2

2 Legal reasons prevent from charging farmers that receive milk quotas out of the state reserve. The state reserve

is filled for instance when farmers cancel renting contract that have been closed before 2000. Such contracts

were valid after the introduction of the quota exchanges and the banning of renting quotas. They could also be

prolonged with the same parties. However, when one of the parties cancelled the contract, the quota cannot be

rented to another party. If then sold through the respective exchange, one third of the quota is claimed by the

state and goes into the state reserve.

4

To illustrate the price setting and transaction rules on German milk quota exchanges, we use

the following example with four bids and four asks. On a regular double Auction (k-multiple

unit double auction) the equilibrium price is set in the range between:

{ } ( ) { }+ += + −*

1 1min , 1 max ,m n n mp k b a k a b with +

+ +≤ > ∈Ν1 1

, and ,n m n ma b a b n m (1)

[ ]* : Equilibrium price; 0,1 ; : Bids ordered descending; : Asks ordered ascending.p k b a∈

Figure 1: Illustration of the k-multiple unit double auction and the pricing rule on German milk quota exchanges

Legend: P*/Q*: Equilibrium for a k-multiple unit double auction. PMB: Preliminary price.

PMB*: Final price. ES: Excess supply. ED: Excess Demand. A: Ask. B: Bid with prices and

volumes in parentheses.

Source: Own.

For a SBDA k would be 0 and thereby the single price would be the lower bound of the range

marked as p* in Figure 1. The equilibrium quantity for a SBDA is Q* which is always at the

intersection of the bid functions.3 In our example in Figure 1 P* is 60 and Q* is 60. On

German milk quota exchanges the preliminary price is set at the minimum excess demand

which is at a price of 50. The equilibrium quantity would still be 60 and the excess demand

would have to be filled out of the state reserve. However, in this case the bid A exceeds the

preliminary price by more than 40 percent; thus, bid A is not considered for the final price

equilibrium. Graphically, all other bids move to the left till B reaches the y-axis. The final

price is set at 40. Bid D is now newly considered; the equilibrium demand is now at 80 and

3 Bidding functions generally characterize the relationship between bids and the underlying true valuations. The

ordered bids are sometimes called the same. To distinguish the two terms here, we call the ordered bids a bid

function.

Q*= QMB

P

Q

ES*

P*

E (100/20)

F (70/20)

G (65/20)

H (49/40)

A (20/20)

B (40/30)

C (60/20)

D (75/40)

PMB

EDMB*

PMB*

QMB*

20

40

60

100

70

65

75

50

0 10 20 30 40 50 60 70 80 90 100

5

supply is as before at 50. The excess demand of 30 is taken from the state reserve and

distributed between the successful bidders with the same share according to their bid volume.

Bidder B for instance bids for 20 units. As the relative excess demand accounts for 37.5

percent of the total volume of successful bids, every bidder receives 37.5 percent of the bid

volume from the state reserve free of charge. Thereby the net price for bidders drops to 25.

Both regulation lead to lower prices and trade (supply) volumes. Trade losses occur by the

reduction in supply and the exclusion of high value bids.

Experimental Design

As many goods and factors in the real world milk quotas have attributes of both private and

common values. However, in modeling different regulations on German milk quota

exchanges we assume private values to be the dominant factor here. This view is support by

the argument “valuation of milk quotas is highly dependent on farm specific characteristics

including production costs, capacity constraints etc.“ (Bogetoft et al. (2003): 197). The studies

undertaken so far in this particular field have to our knowledge followed this approach

(Doyon et al., 2008; Bogetoft et al., 2003; Brümmer and Loy, 2003). The German milk quota

exchanges limit or ban any trading after the auction has closed. Thus, bidders in the

experiment always have only one shot for finding their optimal outcome in each round in the

respective setting, the auction is non sequential. To investigate the major interventions on

German milk quota exchanges we run three treatments. First, the reference treatment is the

SBDA illustrated in Figure 1. The equilibrium price is determined according to equation 1 and

k is equal to zero, the excess demand or supply is found at the intersection of the bidding

functions. To ensure the trade equilibrium the bids with the lowest priority are used to

equalize the market excess. Second, we model the impact of the state reserve policy by setting

equilibrium prices at the minimum excess demand which is then filled by the state reserve.

Every bidder receives according to his/her bid volume the same share out of the state reserve.

The share is determined by the relative excess demand which equals on minus the total supply

divided by the total demand of successful bidders. In the example in Figure 1 every bidder

gets 37.5 percent of their bid volume for free out of the state reserve. Third, we add a price

band for bids to the reference treatment. We calculate a preliminary price according to the

SBDA rule and check whether some bids exceed this price by more than 40 percent. Those

bids are deleted and the price is determined by the same rule again. The price band is applied

only once. Thus, the price determined after deleting bids is always the final price.

For the three treatments we ran lab experiments with students at the Faculty of Agriculture in

their first year. The experiments were promoted during lecture time and interested students

could sign up for participation for one treatment and session only. Each experiment is run

with 16 students. 6 students are assigned to be selling quota and 10 to be buying quota.

Individuals were assigned the role of buyer or seller at random by drawing the number of the

working station. Every student worked at a computerized workstation to receive and submit

information during the experiment. For each scenario five repetitions (sessions) with different

participants were run. In each session, the experiment and the scenario were briefly explained

before written instructions were handed out to every participant. The instruction could be

studied for 15 minutes and questions could be asked and were answered by the instructor

before the start of the session. The experiment then began with three trial runs in which

students tested the functioning of the procedure. In each round every participant receives a

willingness to pay or sell and the volume wanted to buy or sell on screen. Bids and asks are

drawn from random distributions. Prices come from an equal distribution between 0 and 100.

Volumes are drawn from equal distributions between 1 and 9 for buyers and between 1 and 19

for sellers. We choose a rather simple distribution to make calculations and the process of

expectations easier to grasp for participants. After receiving their valuations participants place

their bids/asks. Afterwards they receive results for the equilibrium prices and the information

6

whether or not their bids/asks were accepted on the screen. For the successful bidders the

account information is recalculated by the amount of “money” that was made or lost through

the trade. After three trial rounds the experiment is started over again for the 28 rounds of the

real experiments. In the end the account balance of each participant is taken to calculate the

expense allowance that was paid in cash after the session. On average the entire sessions

lasted about one hour and the average expense allowance was about 10 Euros per participant.

For treatment 2 and 3 we ran 5 sessions, treatment 1 was repeated 4 times.

Optimal bidding

The reference scenario used here is a SBDA which is similar to the P/Q mechanism originally

introduced by Shubik (1977). For this mechanism various properties can be shown such that

each competitive equilibrium (CE) has corresponding to it a Nash-Equilibrium and thereby is

called incentive efficient (Wilson, 1985). However, the competitive equilibrium is not a

natural outcome of the game, in particular when it is non sequential. Smith et al (1982). for

instance come to the conclusion that the P/Q mechanism converges close to the CE, but might

be better described by a monopoly-monopsony equilibrium in which buyers and sellers

withhold their demand and supply respectively to manipulate the price is their preferred

direction. The efficiency is lower than in regular oral double auction; however, this result

changes with more experienced subjects in the experiments (Smith et al, 1982) Overall,

convergence to CE for the oral double auction is reached quickly after some periods of trading

(Plott, 1982). For the P/Q mechanism strategic bids converge to efficient bids as the number

of bidders grows with a rate of convergence that is close to one by the number of bidders

(Gong and Preston McAfee, 1996; Satterthwaite and Williams, 1989). That implies that the

one shot equilibrium might systematically as well as randomly deviate from the CE.4

Nonetheless, we will employ this scenario for comparison, as it is very similar to the

mechanisms studied, it is an operable alternative, and is likely to be more efficient than the

designs used in Treatments 2 and 3. We also use the “deterministic” CE as a second

reference.5 Truth telling would be a dominant strategy, when participants act as price takers

(Nautz, 1995). Starting with this assumption we derive the optimal bidding rule for the second

treatment. In a competitive bidding environment, seller’s bids should not be affected by the

rules in the Treatments 2 and 3. Buyers on the contrary receive a certain share ω of quota out

of the state reserves free of charge; thus, bid prices should increase by 1

1 ω− to reflect the full

valuation for the milk quota bought through the exchange which in only part of the quota

received in total. As valuations and volumes are drawn from random distributions, bidders

have to calculate an expected value forω . An initial guess might be derived as following: The average bid is 5 units and so might be the average excess demand. Supply and demand

intersect at half of average total demand and supply which is between 25 and 30 (Figure 2).

Thus, the relative excess demand might be between 16 to 20 percent. Though this estimation

is not consistent, simulations indicate that the relative excess demand is about 20 percent. In

Table 1, average results for 100.000 repetitions are shown.

Optimal bidding in Treatment 3 is again only relevant to buyers. Considering that the average

equilibrium price might be close to 40 ct (Figure 2), a first rule might be that buyers restrict

their bids to 40 times 1.4 (64 ct). Thus, for instance from 60 ct on the bidding functions stay

constant.

4 “Since everything from electrons, stars and rats to people yield state measure subject to “error” or random

variations the deterministic criterion (CE) can have no serious scientific standing.” (Smith et al., 1982: 201). 5 In the CE all bidders truthfully reveal their valuations.

7

Table 1: Simulation of Treatment 1, 2 and 3

Legend: Excess: In Treatment 1: Absolute excess demand or supply / total demand. In

Treatment 2: Excess demand / total demand. Welfare B(A): Trade gains of all buyers/sellers.

Source: Own.

Figure 2: Expected bids and asks in the simulation

Price

Volume

Source: Own.

Empirical Results

In Table 2 we summarize the average behavior and outcome over all rounds and for each

treatment in the experiment. Additionally, we differentiate between the initial and the final

rounds to estimate the impact of learning or bidding strategy improving. In the reference

SBDA scenario (Treatment 1), bidders perform significant bid shading. On the supply side the

willingness to sell (WPs) is overstated by average asks (Ask) by about 6 percent and on the

demand side average bids (Bid) undermine the willingness to pay (WPb) by also 6 percent

(Table 2). The bid shading predominantly occurs on the supply side at low valuations and on

the demand side at high valuations (Figure 3). Also the bid volumes are lower with respect to

the average given demand and supply (Qs-AskQ/Qb-BidQ). The reduction in supply is higher

than in demand even in relative terms. Smith et al. (1982) describe this behavior as an attempt

Price Volume Excess Welfare B Welfare A Total Welfare

ct t % ct ct ct

Treatment 1 (SBDA) 43.83 26.67 0.23 758.76 450.16 1208.92

Treatment 2 without state reserve 40.48 23.06 0.20 692.59 384.10 1076.69

Treatment 2 43.92 25.42 0.19 974.53 464.03 1438.56

Treatment 3 23.69 13.36 0.41 492.14 176.42 668.55

0 5 10 15 20 25 30 35 40 45 50 55 60

10

20

30

40

50

60

70

80

90

100

8

to exercise market power. As there are only 6 participants on the supply side and 10 on the

demand side, the stronger reduction on supply side appears to be rational. In comparison to

Treatment 1 the supply side strategy is not affected by the design of Treatment 2 and 3. This

is reflected by almost the same extent of bid shading in volumes under Treatment 2 and 3 (Qs-

AskQ). Average asks on the contrary have changed and reflect now almost their true

valuations (WPs-Ask). Their slight deviation mirrors the direction that is observed for the

demand side. The magnitude is significantly smaller which might indicate that a few bidders

have applied the demand strategy to the supply. On the demand side the bid reduction in

volume is slightly higher on average in Treatment 2 and 3. Bid prices, however, change

significantly. In Treatment 2 (state reserve) as bidders expect to receive quota free of charge,

average bids overstate the willingness to pay. In comparison to Treatment 1 the change is

about 8 percent. In Treatment 3 bidders expect to be punished for high bids; thus the bid

shading becomes more pronounced by about 10 percent compared with Treatment 1. On

average bids are 8 ct lower that the true valuation of bidders. Due to the regulations in

Treatment 2 and 3, trade prices and volume decrease with respect to the reference scenario.

Prices go down by 8 (12) percent in Treatment 2 (3); volumes decrease by 12 (32) percent. In

Treatment 2 the decrease in trade volumes equal the reductions in supply. Because of the state

reserve the average total demand transferred to bidders is about 5 units or 20 percent higher

than the supply. This was about to be expected by the simulations runs presented in the

previous chapter. Therefore, net prices decrease by also 20 percent from 43 to 35 ct on

average. The changes in the bidding behavior over time (session rounds) are derived from

results for the first 10 and the final 10 rounds in the respective sessions and treatments. In

Treatment 1 the observed bid shading on the supply side increases for prices and decreases for

volumes both significantly. Auctions average prices decrease and the trade volume increase

which indicate a more competitive behavior. Though bid shading in prices increases, its

impact on equilibrium might not be as severe because it occurs predominantly at high

valuations (Figure 3). Though in absolute smaller terms, the same developments are observed

for the demand side bidders. In the final round bid volumes are close to the true demand

quantities.

Figure 3: Bid functions for Treatment 1

Legend: WP: Willingness to sell/pay in ct/kg; Ask/ Bid: Bids in ct/kg; Results for Treatment

one in Session 1.

Source: Own.

In Treatment 2 we only observe a change in the price bidding functions. The bid shading on

the supply side slightly increases, while is strongly changes on the demand side. Due to the

state reserve policy bidders start with almost true valuations on average (First 10 rounds) but

020

40

60

80

100

Bid

0 20 40 60 80 100WP

020

40

60

80

100

Ask

0 20 40 60 80 100WP

9

end with a significant over valuation or negative bid shading (Final 10 Rounds). Bids are in

the end about 6 ct higher than under Treatment 1 which corresponds to about 12 percent of the

average valuation of bidders. That is still below the theoretical expectation of about 20

percent. Considering a continuing rate of change with experimental repetitions, the expected

percentage would have been reached in the next ten or twenty rounds. Trade volumes in the

final rounds are higher and auction prices are lower compared to the first rounds; adjustments

in the bidding function enable regaining of trade profits compared to SBDA mechanism.

Table 2: Average simulation results for Treatments 1 to 3

Legend: WPs/ WPb: Willingness to sell/pay in ct/kg; Ask/ Bid: Bids in ct/kg; Qs-AskQ/Qb-

BidQ: Difference between the true and revealed demand/supply. Price: Auction price; Trade:

Auction trade volume (In Treatment 2 trade is equal to the supply volume, the excess demand

is covered by the state reserve).

Source: Own.

Treatment 1 Treatment 2 Treatment 3

P/Q State Reserve Price Band

All Rounds

WPs 50.86 49.20 51.09

Ask 53.80 49.87 50.53

WPs-Ask -2.94 -0.66 0.56

Qs-AskQ 3.07 3.08 3.28

WPb 50.55 50.65 50.20

Bid 47.41 51.71 42.11

WPb-Bid 3.14 -1.07 8.09

Qb-BidQ 0.84 1.43 1.44

Price 47.19 43.40 41.46

Trade 27.85 24.68 19.01

First 10 Rounds

WPs 52.98 49.71 51.55

Ask 53.87 50.11 50.58

WPs-Ask -0.88 -0.40 0.97

Qs-AskQ 3.51 3.18 3.37

WPb 50.70 51.42 50.92

Bid 47.86 51.34 41.82

WPb-Bid 2.84 0.08 9.10

Qb-BidQ 1.01 1.37 1.35

Price 47.87 44.80 41.44

Trade 26.23 23.48 19.00

Last 10 Rounds

WPs 48.26 49.55 51.02

Ask 53.47 50.42 50.55

WPs-Ask -5.21 -0.87 0.47

Qs-AskQ 2.34 3.03 3.07

WPb 50.18 49.78 49.61

Bid 46.88 52.62 42.46

WPb-Bid 3.30 -2.84 7.15

Qb-BidQ 0.65 1.41 1.53

Price 46.17 42.96 41.74

Trade 29.17 24.02 19.80

10

Treatment 2 shows bid volumes on the supply and demand side very close between the first

and the final rounds. The bid shading on the supply side is significantly reduced at a relatively

low level. On the demand side the bid shading starts at a very high level. On average bids are

9 ct or almost 20 percent below the true valuations. In the end the shading of bids drops to 7 ct

on average. The reason might be that the estimation of the optimal bidding rule under this

treatment is much more complex; thus, the reduction in bid shading might either a

readjustment or is just another trial in order to match the optimum. Compared to the intuitive

rule that all bids above a certain level are constant, the empirical bid function shows a

considerable amount of high price bids. This is likely inefficient (Figure 4). Also many bids

with same underlying valuations vary significantly indicating that either no clear bidding

strategy is applied among participants or the strategies differ between them.

In Treatment 3 the impact of price bands on the average price is about 2 ct or 4 percent of the

average equilibrium price if we compare the results for the same bids with and without price

bands. That means that even under the adjustment of bidding behavior a considerable amount

of bids is deleted. These numbers and also the trade volume do not seem to improve over

time; thus, there is likely no learning to be detected under this treatment.

Figure 4: Bid functions for Treatment 3

Legend: WP: Willingness to sell/pay in ct/kg; Ask/ Bid: Bids in ct/kg; Results for Treatment

three in Session 5.

Source: Own.

Conclusions

In the paper we design three different treatments to model and analyze the main features of

German milk quota exchanges. The base treatment is a single price multiple unit seller’s

sealed bid double auction which serves as a reference scenario to measure the impact of

alternative regulations. The main features of the German milk quota exchanges are the state

reserves that cover excess demands free of charge and the price bands that limit bids on the

demand side. Both regulations aim at lowering equilibrium prices for quota buyers.

We employed first year students in Agricultural Sciences at the University of Kiel to run the

three different treatments. For each treatment 16 participants randomly assigned to 6 sellers

and 10 buyers trade under the respective market designs. Incentives are given by trade profit

based payments for each participant.

Experimental results show that both regulations have a significant impact on equilibrium

prices and trade volumes. In both designs the market outcomes (prices and volumes)

significantly decrease compared to a regular double auction. This result does not hold for the

020

40

60

80

100

Bid

0 20 40 60 80 100WP

020

40

60

80

100

Ask

0 20 40 60 80 100WP

11

demand in Treatment 2 (state reserve). When the volume transferred by the state reserve is

considered the total demand is higher than for the regular seller’s bid double auction. In

particular the price band leads to a significant reduction in trade of about 30 percent. Thus,

lower quota prices come at a significant cost of reduced trade gains.

The experiments also show that participants do not efficiently adjust their bidding functions to

the mechanisms in the treatments. Bid shading appears in the regular double auctions on both

market sides and for prices and volumes. Though the supply side bid function is not affected

by the treatments, a reduction in bid volume occurs in all treatments. The bidding price

functions deviate slightly in the direction that is expected for the demand side which is not

necessarily efficient for the demand side. On the demand side the expected direction of

adjustments occurs. Bids are increased when state reserves cover the excess demand and

reduced when price bands are applied. However, at least for the state reserve treatment the

optimal adjustment is not accomplished. In all scenarios a similar bidding function applied by

all bidders is missing which in addition is responsible for the observed inefficiencies. The

trade losses that might be caused by randomness and/or other systematic factors could likely

be corrected by continued trading. Also in the real world, the opening of at least alternative

markets and transaction mechanisms might lead to improved trade performances.

References

Bogetoft P., K.Nielsen and H.Olesen (2003): Single Bid Restriction in Milk Quota

Exchanges, European Review of Agricultural Economics, 30, pp. 193-215.

Brümmer B., J.-P. Loy und C. Struve (2003): Beurteilung des Milchquotenbörsensystems in

Deutschland. Agrarwirtschaft 52(6): 275-288.

Doyon, M., V. Simard, et al. (2008): An Experimental Analysis of Modifications to the

Centralized Milk Quota Exchange System in Quebec. Canadian Journal of Agricultural

Economics/Revue canadienne d'agroeconomie 56(3): 295-312.

Gong J. and R. Preston McAfee (1996): Convergence to Efficiency in Double Auctions.

Advances in Applied Micro-Economics, Volume 6, ed: Michael Baye, Greenwich, CT: JAI

Press: 63-96.

McAfee, R. P. (1992): A dominant strategy double auction. Journal of Economic Theory

56(2): 434-450.

Nautz D. (1995): Optimal bidding in multi-unit auctions with many bidders. Economics

Letters, Volume 48, Issues 3-4: 301-306.

Nautz D. and E. Wolfstetter (1997): Bid shading and risk aversion in multi-unit auctions with

many bidders, Economics Letters, Volume 56, Issue 2: 195-200.

Plott C. R. (1982): Industrial Organization Theory and Experimental Economics. Journal of

Economic Literature 20(4): 1485-1527.

Satterthwaite, M. A. and S. R. Williams (1989): The Rate of Convergence to Efficiency in the

Buyer's Bid Double Auction as the Market Becomes Large. The Review of Economic Studies

56(4): 477-498.

Shubik M. (1977): A Price-Quantity Buy-Sell Market With and Without Contingent Bids,

Cowles Foundation Discussion Paper 455, May 16, 1977.

Vernon L. Smith, Arlington W. Williams, W. Kenneth Bratton, Michael G. Vannoni (1982):

Competitive Market Institutions: Double Auctions vs. Sealed Bid-Offer Auctions. The

American Economic Review, Vol. 72, No. 1 (Mar., 1982), pp. 58-77

Wilson R. (1985): Incentive Efficiency of Double Auctions. Econometrica 53(5): 1101-1115.